I’m building a split keyboard from scratch. It’s not finished — one side is soldered, the brain isn’t wired up yet, and the finish is still on the to-do list. But it’s far enough along to be worth writing about, including the trick I’m most pleased with.
The design
I started from the layout of a commercial split keyboard I liked, then modified it. Each half got an extra row of keys — six more per side — plus room for a battery, a small OLED display and a rotary knob, none of which the original had. The knob isn’t wired or programmed yet; the display is still a plan. But the space is there, which is the important part at this stage.
The plate is 3D printed on my Creality Ender-3 V3 KE, with holes for the keys and mountings for the hot-swap sockets. It’s a tight fit — deliberately so. Once all the wiring is in, those sockets aren’t going anywhere.
Soldering
Thirty-two keys on one side: hot-swap sockets and per-key LEDs, all hand soldered. If you’ve never soldered a keyboard matrix, the tedium is real, but there’s something satisfying about a row of clean joints.
Here’s the trick: before moving on, I wired an ESP32 dev board up as a test jig and drove every LED in the circuit. Any bad joint or break shows up immediately as a dark LED. It took minutes and confirmed the whole side was good — far better than discovering a cold joint after the board is assembled and the other side is done. If you’re hand-wiring anything with LEDs, test in circuit as you go. Future you will be grateful.
The brain (planned)
Key scanning will run on an nRF52840 with ZMK — the standard choice for wireless split boards, and what my firmware config already targets. The more interesting idea is a second chip: an ESP32-C6 as a display coprocessor, talking to the nRF over UART with tiny state messages (LAYER:1, CAPS:ON, that sort of thing). The nRF owns the keyboard; the ESP32 owns the screen. I’ve written up the design reasoning separately — the short version is that the interface between the chips matters more than the chips themselves, and I’m still validating whether ZMK’s own display support makes the coprocessor unnecessary.
What’s left
Honestly, a fair bit: hook up the nRF, wire the battery, OLED and knob, program the knob, then do the second half. The plate bowed slightly — probably heat during printing — but it’ll pull flat once screwed into the enclosure.
Then the finish. I want the surface to look professional, so I’ve got Bondo and a good primer waiting. Fill, sand, prime, paint. That’s a weekend on its own.
I have all the materials. The only thing I need is time.
As an Amazon Associate, Headless Diaries earns from qualifying purchases made through links on this page.